Asymptotic expansions of some Toeplitz determinants via the topological recursion (Q2300916)

Consider the random matrix $M_{n}$ given by the circular ensemble on the complex numbers; this is in fact the Haar measure on the set of unitary matrices with complex entries. Then $M_{n}$ has with probability one $n$ eigenvalues of the form $e^{i\theta_{j}}$, with $\theta_{j}\in [0,2\pi],\ j=1,2,\ldots,n$. Now let $\mathcal{I}\subset [-\pi,\pi]$. The author wants to estimate {the probability that all values $\theta_{j}$ lie inside $\mathcal{I}$}, when $n$ gets large. In other words, for particular sets $\mathcal{I}$, he provides asymptotic formulas for $$Z_{n}(\mathcal{I}):=E[f(\theta_{1})\ldots f(\theta_{n})],$$ where $f$ is the indicator functions of $\mathcal{I}$. The particular sets are of the form $\mathcal{I}:=\bigcup_{j=1}^{d}[\alpha_{j},\beta_{j}]$. It is known that $Z_{n}(\mathcal{I})$ can be calculated by means of the determinant of a certain Toeplitz matrix, and also using a Hermitian matrix with a potential $V(x):=\log(1+x^{2}),\ x\in \mathbb{R}$. These reformulations help to carry out the analysis. To achieve the task, the author uses some correlation functions that fit into a concept called topological recursion which originally comes from the study of meromorphic functions on Riemann surfaces. Some recursive equations for the correlation functions can be found and used to ultimately give the asymptotics of $Z_{n}(\mathcal{I})$. When $d=1$ and $\beta_{1}=-\alpha_{1}=\epsilon \pi $, the asymptotics were found by \textit{H. Widom} [Adv. Math. 13, 284--322 (1974; Zbl 0281.47018)]. The cases that the article is able to tackle are the following: \begin{itemize} \item $d=1$ for general $\beta_{1}$ and $\alpha_{1}$.\\ \item $d=2r+1\geq 3$ odd, for $-r\leq j\leq r$, \[ [\alpha_{j}, \beta_{j}]=\left[ \frac{2\pi j}{d}-\frac{\pi \epsilon}{d}, \frac{2\pi j}{d}+\frac{\pi \epsilon}{d}\right]. \] \item $d=2s$ even, for $-(s-1)\leq j\leq s$, \[ [\alpha_{j}, \beta_{j}]=\left[ \frac{\pi (j-\frac{1}{2})}{s}- \frac{\pi \epsilon}{d}, \frac{\pi (j-\frac{1}{2})}{s}+ \frac{\pi \epsilon}{d} \right]. \] \end{itemize}